Bookmark and Share

Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

re: Re: st: ddens option for McCrary density test in Nichols' rd command

From   "Ariel Linden, DrPH" <[email protected]>
To   <[email protected]>
Subject   re: Re: st: ddens option for McCrary density test in Nichols' rd command
Date   Tue, 13 Sep 2011 11:02:29 -0400

Hi Austin,

I actually like the ddens option, so I wouldn't necessarily delete it. You
made it perfectly clear in the help file that this is an ad-hoc approach,
and you do provide the link to McCrary's site.

McCrary states that his code will only be available on his site for a
limited time, so it is likely that future RD enthusiasts will have to
recover it from another location. 

Perhaps you could incorporate his code in rd rather than using ddens? This
will allow for one-stop-shopping (as you did with optimal bandwidth

Date: Mon, 12 Sep 2011 10:32:35 -0400
From: Austin Nichols <[email protected]>
Subject: Re: st: ddens option for McCrary density test in Nichols' rd

Jen Zhen <[email protected]> :
I guess I should just remove that option to clear up the confusion.
- -ddens- was an ad hoc placeholder to remind people to do McCrary's
test, which the help file has always linked to.  The code to compute a
difference in kernel densities is largely untested and does not
provide SEs, and was intended only for bootstrapping, for which there
is no proof of consistency, so it is unsuited to the newer versions of
- -rd-.

On Mon, Sep 12, 2011 at 9:57 AM, Jen Zhen <[email protected]> wrote:
> Right, apologies for being imprecise on this. Remains the problem
> though that it does not implement any other density test either.
> On Mon, Sep 12, 2011 at 3:33 PM, Nick Cox <[email protected]> wrote:
>> The help file for -rd- (SSC; please explain provenance!!!) says
>> ddens requests a computation of a discontinuity in the density of Z.
This is
>>    computed in a relatively ad hoc way, and should be redone using
>>    test described at
>> I read that as clearly stating that this option does not implement
>> McCrary's test.
>> Nick
>> On Mon, Sep 12, 2011 at 2:24 PM, Jen Zhen <[email protected]> wrote:
>>> Dear Statalist members,
>>> when I use the -ddens- option with Austin Nichols' -rd- command, I
>>> just get the same main estimates output as without that option,
>>> instead of a computation of McCrary's density test. Is it possible
>>> that this option is not working as it should, or is there anything
>>> particular I need to take into account when using this?

*   For searches and help try:

© Copyright 1996–2018 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   Site index